Data collection for this study was done during the first week of March 2013. Prior to collecting the data, a letter of authorisation (Appendix 1) to collect data was written to the schools’ principals in order for them to grant permission for this study to be carried out. After authorisation was granted, class teachers were asked to assist with the random selection of learners for the study as well as with the supervision of the task. The participants completed this task in the classroom. The hand written scripts were collected and then an experienced English lecturer who is a native speaker of English was consulted to assist with the identification of the ungrammatical sentences.
This was to enable a thorough and detailed analysis given the fact that I am not an English native speaker and only have six years of experience in language teaching. This process involved underlining the non-standard forms and next to them providing the correct version.
After the errors were identified, I coded them using the marking codes shown in Appendix 4. I then classified the errors into categories as can be seen in chapter 4. The West Virginia assessment rubric (The West Virginia Education Department, 2011) was used in the identification of discourse errors. This rubric was designed by the Education Department of West Virginia with the objective of guiding English language teachers in the development, improvement, and evaluation of learners’ writing, in order to provide valuable feedback that will help in language learning. I have chosen this rubric because it comprehensively covers five key language components, aligned to grade level objectives, and evaluates the analytic traits of organisation, development, sentence structure, word choice and grammar usage, and punctuation. From my observation, research on Error Analysis neglects the use of assessment rubrics leaving out a great deal of errors relating to discourse. Having compared this rubric to the Namibian rubric that is currently used in schools, I
58 felt it is more comprehensive to use for the analysis of the data for this study (see chapter 4, section 4.6).
Numerous studies (Johnson and Johnson, 2008; Kesharvarz, 2006; Ellis, 2003; Cherrington, 2004;
Ferris, 2002; James, 1998 and Corder, 1974) on Error Analysis informed the process used to analyse data for this study. These studies provided guidelines with regard to the use of Corder’s (1981; 1967) model for Error Analyses, in that the following steps were followed:
Firstly, the 100 written scripts were read. The norm is the target language and any deviation from the norm was viewed as an error (Richards and Schmidt, 2002). A word or sentence was taken as a basic unit of analysis and then erroneous language elements were coded. Corder’s (1981; 1967) model (chapter 2, figure 2) for error identification was used in this study. This model advises that the erroneous sentence should be translated verbatim into the learner’s first language in order to determine the transferred elements. The sentence is then reconstructed in the target language;
where an error was identified, it was coded. Given the fact that I am not a native speaker of English, the error identification and reconstruction of the sentence in the target language was done by an experienced English native speaker. This was to ensure a thorough identification of errors.
Hobson (1999) pointed out that determining a norm is problematic because it depends on the linguistic context and the relation between the speaker and hearer. For instance, covert errors pose difficulties because the analyst tries to identify what the learner intended to say. An example is given below:
(a) I was the first in race.
The analyst might assume that the learner wanted to say “I was the first to race”, while he/she meant that she/he came first in the race. This shows that future research should consider both written and spoken language so that learners can clarify what they mean.
The next step after the identification of errors was the description and classification of errors into types. The purpose of providing a description of errors, according to James (1998), is to make the errors explicit. It is indispensable for counting errors and it is a basis for creating categories since it reveals which errors are different or the same. The learner’s language had to be described in
59 terms of a language system. The Interlanguage hypothesis suggests that “learner language” is a language in its own right and should not be described in terms of the target language (James, 1998).
Despite James’s (ibid) argument, a comparative technique is important in tracing whether errors are caused by first language interference. This is to ensure that appropriate remedial methods are designed to help learners learn their second language.
Ellis (1994) and Krashen (1982) proposed that “surface strategy taxonomy” is required to describe errors (see chapter 2.5.3 for definition). The surface structures are altered by means of operations such as:
(a) Omission, the absence of an item that must appear in a well formed sentence (I happy - I am happy);
(b) Addition, the presence of an item that must not appear in a well formed sentence (The dog it can bark - The dog can bark);
(c) Misordering, the incorrect placement of a morpheme or group of morphemes in a sentence (She is going not to the hospital - She is not going to the hospital).
Errors were classified into various categories based on their nature. Kesharvarz’s (2006) and Corder’s (1967) linguistic error taxonomy was used to categorise errors collected in this study.
This taxonomy allows errors to be classified according to the language component or linguistic constituent which is affected by errors. Amongst the language components as highlighted in chapter 2 are phonology/orthography errors, lexico-semantic, morpho-syntax, discourse and technique error categories.
The linguistic error taxonomy is used in the study of Error Analysis as it is found to be useful for organising the collected data (Kesharvarz, 2006; Llomaki, 2005; Mohamed, 2004 and Keiko, 2003). In addition, the classification of errors into various categories is beneficial to teachers and curriculum developers because teaching and learning can be ordered step by step from simple to complex structures of the language. After categorising each error, I quantified the frequency of occurrence of different types of errors. The results were tabulated and are illustrated graphically (chapter 4). The descriptive statistics for the results are provided, including the frequencies of errors, error types and percentages. Some of the erroneous sentences (as will be seen in chapter 4) have been double coded because they consist of more than one error.
60 The last step was to provide probable explanations as to the causes of errors that the learners made.
Corder’s (1981) model for error explanation was used in order to determine whether or not the errors are a result of first language interference. This involved making a well formed reconstruction of the learner response in English, comparing the reconstructed sentence with the original idiosyncratic sentence, checking for the difference between the original and reconstructed sentence, translating verbatim the sentence to Oshiwambo to see if it is a plausible interpretation in the given context and then translate the sentence in Oshiwambo back to English in order to provide a reconstructed sentence (see chapter 4 for examples). During this analysis comparisons were made between Oshiwambo and English language structure to determine the transferred language items from Oshiwambo to English. Errors were then labeled as either resulting from first language interference, being interlingual errors or overgeneralisation of the target language rule, or lack of knowledge of the second language, being intralingual errors (Kesharvarz, 2012; Ellis, 2003 and James, 1998). Deciding on whether the error is intralingual or interlingual is a bit problematic because a spelling error may belong to both categories. This is discussed in chapter 4 of this thesis.